AlphaCrafter: A Full-Stack Multi-Agent Framework for Cross-Sectional Quantitative Trading
Researchers have introduced AlphaCrafter, a novel full-stack multi-agent framework designed to address the challenges of non-stationary financial markets in quantitative trading. Unlike traditional approaches that treat factor discovery and execution as static or isolated processes, AlphaCrafter employs a continuously adaptive pipeline driven by three specialized agents. The Miner agent utilizes Large Language Models to expand the factor pool, the Screener assesses market conditions to construct regime-conditioned factor ensembles, and the Trader executes strategies under explicit risk constraints. This closed-loop system aims to eliminate behavioral noise and manual intervention, ensuring systematic rationality. Extensive experiments conducted on the CSI 300 and S&P 500 indices demonstrate that AlphaCrafter consistently outperforms state-of-the-art baselines in risk-adjusted returns while exhibiting the lowest cross-trial variance. The study confirms that an integrated, adaptive design from factor discovery to execution yields robust trading performance, marking a significant advancement in automated, rationality-driven quantitative finance systems.
Wire timeline
AlphaCrafter: A Full-Stack Multi-Agent Framework for Cross-Sectional Quantitative Trading
Researchers have introduced AlphaCrafter, a novel full-stack multi-agent framework designed to address the challenges of non-stationary financial markets in quantitative trading. Unlike traditional approaches that treat factor discovery and execution as static or isolated processes, AlphaCrafter employs a continuously adaptive pipeline driven by three specialized agents. The Miner agent utilizes Large Language Models to expand the factor pool, the Screener assesses market conditions to construct regime-conditioned factor ensembles, and the Trader executes strategies under explicit risk constraints. This closed-loop system aims to eliminate behavioral noise and manual intervention, ensuring systematic rationality. Extensive experiments conducted on the CSI 300 and S&P 500 indices demonstrate that AlphaCrafter consistently outperforms state-of-the-art baselines in risk-adjusted returns while exhibiting the lowest cross-trial variance. The study confirms that an integrated, adaptive design from factor discovery to execution yields robust trading performance, marking a significant advancement in automated, rationality-driven quantitative finance systems.
cs.AI updates on arXiv.org